Kumo Co-Founder Hema Raghavan Named to Inc.’s 2026 Female Founders 500

Learn more
9Regression · Workforce Planning

Branch Demand Forecasting

How many tellers does each branch need next week?

Book a demo and get a free trial of the full platform: data science agent, fine-tune capabilities, and forward-deployed engineer support.

By submitting, you accept the Terms and Privacy Policy.

Loved by data scientists, ML engineers & CXOs at

Catalina Logo

A real-world example

How many tellers does each branch need next week?

US banks operate 70,000+ branches, and labor is the largest controllable cost at 55-65% of branch operating expense (BAI). Overstaffing wastes $15-25K per branch annually in idle teller hours, while understaffing drives average wait times above 8 minutes, directly correlating with a 12% drop in customer satisfaction (J.D. Power). The challenge: branch traffic depends on payroll cycles, local events, weather, nearby business hours, and even competitor branch closures. Static scheduling models based on last year's averages miss these dynamic signals.

How KumoRFM solves this

Relational intelligence built for banking and financial data

Kumo connects branch profiles, historical foot traffic, transaction volumes, local event calendars, weather data, and regional payroll cycles into a relational graph. The model predicts that Branch S-14 will see 340 transactions next Tuesday because it is the first of the month (payroll deposits), a local employer just switched to bi-weekly pay, and a competitor branch 2 miles away recently closed. These cross-table signals produce staffing forecasts 35% more accurate than static models.

From data to predictions

See the full pipeline in action

Connect your tables, write a PQL query, and get predictions with built-in explainability — all in minutes, not months.

1

Your data

The relational tables Kumo learns from

BRANCHES

branch_idnameregionavg_daily_txnsteller_stations
BR-014Union SquareWest2856
BR-022MidtownNortheast4208
BR-037LakesideMidwest1804

DAILY_TRAFFIC

branch_iddatetransactionsavg_wait_mintellers_on_duty
BR-0142025-09-293104.25
BR-0142025-09-303958.75
BR-0222025-09-304806.17

LOCAL_EVENTS

branch_iddateevent_typeexpected_impact
BR-0142025-10-01Month Start (Payroll)High
BR-0142025-10-01Competitor Branch ClosureMedium
BR-0222025-10-03Local FestivalLow

WEATHER_FORECAST

regiondateconditiontemp_fprecipitation
West2025-10-01Sunny720%
Northeast2025-10-01Rain5880%
Midwest2025-10-01Cloudy6520%
2

Write your PQL query

Describe what to predict in 2–3 lines — Kumo handles the rest

PQL
PREDICT SUM(DAILY_TRAFFIC.TRANSACTIONS, 0, 7, days)
FOR EACH BRANCHES.BRANCH_ID
3

Prediction output

Every entity gets a score, updated continuously

BRANCH_IDDATEPREDICTED_TXNSTELLERS_NEEDEDVS_SCHEDULED
BR-0142025-10-013857+2
BR-0142025-10-0226050
BR-0222025-10-013506-1
4

Understand why

Every prediction includes feature attributions — no black boxes

Branch BR-014 (Union Square), Oct 1

Predicted: 385 transactions, 7 tellers needed

Top contributing features

Month-start payroll cycle

1st of month

32% attribution

Competitor branch closure spillover

+45 txns est.

24% attribution

Day-of-week pattern (Wednesday)

Above avg

18% attribution

Weather (sunny, high foot traffic)

Sunny 72F

14% attribution

Regional employment trend

+2.1% YoY

12% attribution

Feature attributions are computed automatically for every prediction. No separate tooling required. Learn more about Kumo explainability

Bottom line: Reduce average wait times by 30% and cut overtime costs by $15-25K per branch annually, translating to $100-175M in savings across a 7,000-branch network.

Topics covered

branch staffing predictionteller demand forecastingbank workforce planning AIbranch traffic predictiongraph neural network workforceKumoRFMbranch operations optimizationrelational deep learning staffingbank branch analyticsstaffing optimization banking

One Platform. One Model. Predict Instantly.

KumoRFM

Relational Foundation Model

Turn structured relational data into predictions in seconds. KumoRFM delivers zero-shot predictions that rival months of traditional data science. No training, feature engineering, or infrastructure required. Just connect your data and start predicting.

For critical use cases, fine-tune KumoRFM on your data using the Kumo platform and Data Science Agent for 30%+ higher accuracy than traditional models.

Book a demo and get a free trial of the full platform: data science agent, fine-tune capabilities, and forward-deployed engineer support.